US20110296353A1 - Method and system implementing user-centric gesture control - Google Patents

Abstract

A user-centric method and system to identify user-made gestures to control a remote device images the user using a three-dimensional image system, and defines at least one user-centric three-dimensional detection zone dynamically sized appropriately for the user, who is free to move about. Images made within the detection zone are compared to a library of stored gestures, and the thus identified gesture is mapped to an appropriate control command signal coupleable to the remote device. The method and system also provides of a first user to hand off control of the remote device to a second user.

Description

RELATION TO CO-PENDING APPLICATION
• Priority is claimed from co-pending U.S. provisional patent application Ser. No. 61/217,355 filed on May 29, 2009, entitled User-Centric Gesture Control.
FIELD OF THE INVENTION
• The invention relates generally to methods and systems usable with a depth imaging system to enable a user to remotely control a device, and more specifically to such methods and systems enabling remote control through recognition of user gestures defined within user-centric four-dimensional (x, y, z, time) space.
BACKGROUND OF THE INVENTION
• Remote control of devices including video devices has evolved from use of IR or acoustic type remote controls held by a user to control television sets and the like equipped with IR or acoustic recognition systems, to imaging systems that attempt to image the user in two or preferably three dimensions to recognize movements or gestures intended to control the television or other device.FIG. 1 depicts a genericprior art system10 in which adevice20, here a television, is remotely controlled by auser30, whose head is shown from the back in the figure.System10 includes at least one imaging system, here a camera, e.g.,40-1,40-2, coupled electronically to asignal processor unit50, whose processor output can control operation oftelevision20.
• The field of view of camera(s)40-1,40-2 encompasses at least a portion of three-dimensional space in which the user can make gestures, for example with at least one hand (e.g., left hand60) to controltelevision20. If conventional RGB or gray scale images are acquired, then typically two spaced-apart cameras40-1,40-2 will be employed. Ideally, allowable gestures would include moving user hand(s) towards or away fromtelevision20, but RGB or gray scale cameras, including a pair of such cameras disposed stereographically, might not correctly discern such movement relative tosystem10. RGB or gray scale cameras are readily confused by ambient lighting including light generated by the television display itself, by the clothing of the user, e.g., a white hand in front of a user's white shirt, by reflectivity of objects within the field of view, etc.
• Various imaging systems that seek to acquire three-dimensional images of a user creating gestures intended to control a device are known in the art. Some three-dimensional imaging systems use so-called parallel techniques and may include two-cameras, such as shown inFIG. 1. Various two-camera implementations include so-called passive stereo in which a sparse depth map is created in which only some sensor pixels in the depth map actually contain depth information. Another two camera approach to acquiring depth images is texture patterned stereo, in which the depth system creates a pattern that generates texture but does not encode depth information. If a speckle-like randomly patterned illumination is used, there may be sufficient texture in the imaged scene to enable creation of a dense depth map. Yet another type of two camera imaging system is depth-coded patterned stereo, in which a patterned Illumination source codes depth information and can provide a dense depth map. A problem common to many two-camera systems is occlusion and so-called correspondence ambiguity. It can be challenging to combine the imagery acquired by two spaced-apart cameras to unambiguously determine depth in an imaged scene.
• Some parallax imaging methods use a single camera with a patterned source of illumination. So-called structured light systems can create a near-far qualitative depth map, but may suffer from an imprecise baseline. PrimeSense, an Israeli company, markets such structured light systems. So-called active stereo single camera systems can acquire a dense depth map with a precise baseline.
• Another and somewhat superior method of three-dimensional imaging uses time-of-flight (TOF) information to create a dense depth map. Canesta, Inc. of Sunnyvale, Calif. (assignee herein) has received several dozen U.S. patents directed to methods and systems that can acquire true depth images. Exemplary such U.S. patents received by Canesta, Inc. include U.S. Pat. No. 6,323,942 (2001) CMOS-Compatible Three-Dimensional Image Sensor IC, U.S. Pat. No. 6,515,740 (2003) Methods for CMOS-Compatible Three-Dimensional Image Sensing Using Quantum Efficiency Modulation, U.S. Pat. No. 6,522,395 (2003) Noise Reduction Techniques Suitable for Three-Dimensional Information Acquirable with CMOS-Compatible Image Sensor ICs, U.S. Pat. No. 6,614,422 (2003) Methods for Enhancing Performance and Data Acquired from Three-Dimensional Image Systems, U.S. Pat. No. 6,674,895 (2004) Methods for Enhancing Performance and Data Acquired from Three-Dimensional Image Systems, U.S. Pat. No. 6,678,039 (2004) Method and System to Enhance Dynamic Range Conversion Useable with CMOS Three-Dimensional Imaging, U.S. Pat. No. 6,710,770 (2004) Quasi-Three-Dimensional Method and Apparatus to Detect and Localize Interaction of User-Object and Virtual Transfer Device, U.S. Pat. No. 6,906,793 (2005) Methods and Devices for Charge Management for Three-Dimensional Sensing, U.S. Pat. No. 7,151,530 (2006) System and Method for Determining an Input Selected by a User Through a Virtual Interface, U.S. Pat. No. 7,176,438 (2007) Method and System to Differentially Enhance Sensor Dynamic Range Using Enhanced Common Mode Reset, U.S. Pat. No. 7,212,663 (2007) Coded-Array Technique for Obtaining Depth and Other Position Information of an Observed Object, U.S. Pat. No. 7,321,111 (2008) Method and System to Enhance Differential Dynamic Range and Signal/Noise in CMOS Range Systems Using Differential Sensors, U.S. Pat. No. 7,340,077 (2008) Gesture Recognition System Using Depth Perceptive Sensors, U.S. Pat. No. 7,352,454 (2008) Methods and Devices for Improved Charge Management for Three-Dimensional and Color Sensing, and U.S. Pat. No. 7,507,947 (2009) Method and System to Differentially Enhance Sensor Dynamic Range Using Enhanced Common Mode Reset.
• Typically a TOF system emits optical energy and determines how long it takes until at least some of that energy is reflected by a target object and arrives back at the system to be detected by an array of pixel detectors. If t₁denotes roundtrip TOF time, then the distance between target object and the TOF system is Z1, where Z1=t₁·C/2, where C is velocity of light. Most Canesta TOF systems are phase-based and compare shift between phase of the modulated emitted optical energy and phase of the reflected energy in determining depth Z. Canesta TOF systems are operable with or without ambient light, have no moving parts, and can be mass produced using CMOS techniques. Phase-based TOF systems are also believed available from PMD Technology of Siegen, Germany, Mesa Imaging, AG of Zurich, Switzerland, and possibly Optrima NV of Brussel, Belgium.
• Another method of TOF systems that does not measure phase shift is the shutter type TOF system. The shutter may be an active optic device, perhaps GaAs as developed by 3DV Corp. of Israel, or perhaps an electronic shutter, e.g., CMOS, as developed by TriDiCam GmbH of Germany.
• Three-dimensional imaging may be accomplished without using a parallel method, or a TOF method, for example by using spaced-apart cameras from whose images relative or inferred depth Z information may be had. Such systems are believed to be developed by XTR 3D Company of Israel. Alternative methods for inferring depth may rely upon camera motion, so-called structure-from-motion analysis, but these methods are not deemed sufficiently fast for use in a gesture recognition system. Other methods for inferring depth include so-called depth-from-focus techniques in which the focal plane of an imaging camera is changed to create a depth map. However such techniques may not be adequately fast or accurate for real-time gesture recognition.
• Having briefly reviewed the various methods known in the art for obtaining depth or Z images, consider now an exemplary prior art approach to gesture recognition with reference toFIG. 1. Assume thatsystem10 include adisplay20 whose characteristic(s) auser30 will attempt to influence or alter using user-made gestures that are imaged here by spaced-apart cameras40-1,40-2.System10 is what may be termed device-centric and typically requires closed-loop visual feedback betweenuser10 and a portion of what is displayed ontelevision system20.FIG. 1 shows, for example, acursor70 near the upper left corner on the television display, and also shows a double-arrow icon80 near the right edge of the television display. In this example, if the user can causecursor70 to move to the right, in the direction ofphantom cursor70′, and overlie the upper or lower portion ofarrow80, the user can thus cause an increase or decrease in the sound volume fromtelevision20.
• In practice,system10 will have pre-defined several gestures that the user will know a priori. For example, to movecursor70 to the right, the user may move the left hand to the right, as indicated by the position ofphantom hand60′. Unfortunately doing so involves hand-eye coordination between the displayed cursor ontelevision20, and the user's hand position, as imaged by cameras40-1,40-2. The (x, y, z) coordinate system relied upon bysystem10 is an absolute coordinate system that is defined relative totelevision set20. This coordinate system means that the distance ΔX′ through which the user's hand must be moved to move the cursor a distance ΔX on the television display is not constant. Thus, if the user is say 8′ (2.5 m) away from the television set, distance ΔX′ will be substantially greater than if the user were say 4′ (1.25 m) away from the television set. In addition to this varying distance sensitivity, the user must keep an eye on the cursor position. In the example ofFIG. 1, once the user moves the cursor to the desired up or down portion ofdouble arrow80, the user might then confirm this selection, perhaps by moving the hand in the direction of the television screen. Having thus executed the desired correction to the television volume,system10 can automatically remove both the cursor and double arrow from the television display. If the user later wishes to make some other adjustment, perhaps to change channels ontelevision20, the user will make some other gesture known tosystem10, and the cursor and other relevant icon(s) or images will appear on the television screen.
• While device-centric systems such as described inFIG. 1 can work, more or less, there is room for improvement. The necessity for hand-eye feedback between the user and what is presented on the television screen may not be desirable for all classes of users. Such feedback requires some user training, e.g., how much hand movement will cause how much screen movement at what distance away from the screen. This need for user training may arise because device-centric systems have variations as the user changes position relative to the device that can affect the user's feel for device control.
• For example, unless scaling is done correctly, an action that requires some subtle motion when the user is far from the device can require large motions when the user is close to the device. Unless addressed in some fashion, this scaled feedback characteristic of some prior art systems can limit the type of user gestures that can reliably be recognized and acted upon.
• Further, systems such as described inFIG. 1 are device-centric in that the three-dimensional coordinate system used by the system is defined relative to the device, and not to the user. In some applications the device-centric nature of the system can result in ambiguous recognition of what device control action was intended by a given user gesture. For example, a user gesture intended to increase the channel number should not be misinterpreted as a user desire to increase the device volume setting, etc.
• What is needed is a remote control method and system that does not require hand-eye feedback between the user and the device being controlled. Preferably such method and system would employ a user-centric relative coordinate system rather than an absolute device-centric coordinate system. Such method and system would free the user from undue concentration upon the device screen to implement remote control. Preferably such method and system should use three-dimensional rather than two-dimensional image sensing, be intuitive to the user, and not require substantial user training. Further such method and system should reliably recognize user gestures without ambiguous interpretations. Gestures should be user-friendly to perform and remember, and should be defined to be unambiguous with good detection discrimination characteristics. Preferably gestures should have no state, e.g., nothing to remember, and should permit transitioning to another gesture unambiguously.
• The present invention provides such a remote control method and system.
SUMMARY OF THE PRESENT INVENTION
• The present invention provides a remote control system and method that is user rather than device centric, and thus relies upon local user-centric coordinates rather than absolute, device or camera system, coordinates. A local environment or three dimensional zone of interaction is defined about the user, and can move as the user moves. The zone of interaction preferably is customized to the user in the sense that a large adult user will have a larger volume zone of interaction than a small child user. The user zone of interaction defines the three-dimensional space in which gestures will be made and detected. Defined user gestures made within the zone of interaction are recognized by a preferably three-dimensional imaging system disposed on or about the device to be controlled. Preferably a small number, three perhaps, three-dimensional hot spots or hot zones are defined within this larger zone of interaction. While user gestures are not confined to be made solely within these hot spots, defining these hot spots enables detection of specific touch gestures. Such user gestures made within a hot spot preferably include at least “touching” the hot spot, and drawing a shape (perhaps a circle or an arc) within the hot spot. As such, the system and method uses real-world coordinates and in terms of feedback can substantially de-couple the user from the device being controlled. Because the local environment is defined relative to the user, a form of metric confirmation exists such that user gestures made near or far from the device are recognized without the need for user hand-eye coordination or user-device calibration. In embodiments of the present invention, preferably there is always at least a coarse one-to-one mapping between the user's local environment zone of interaction and the device display, which mapping is completely transparent to the user.
• A library of user friendly and intuitive gestures is pre-defined such that preferably no substantial user training is required to learn or to remember the gestures. The gestures themselves are defined so as to be unambiguous with good detection discrimination characteristics. The different gestures preferably have no state, e.g., are memory-less, and preferably permit transitioning to another gesture unambiguously. The user need not even look at the device being controlled, e.g., a television, to make a gesture. Further, there is no compelling need for a cursor display, and there is no visual feedback that requires hand-eye coordination. Good gesture recognition according to the present invention relies upon two detection properties, namely how well the three-dimensional imaging system images the gesture, and how well the inventive method can discriminate the gesture from other potential gestures, motions, or noise.
• The system includes a preferably three-dimensional imaging system, such as is known in the art, and a processor unit that includes memory storing at least one software routine that defines a library of user gestures, and at least one algorithm for interpreting data from the three-dimensional imaging system and assigning to the data an appropriate user gesture made within a hot zone in a region of interaction defined relative to the user. The processor unit preferably includes a processor that executes the algorithm and issues appropriate command signals to the device, e.g., television, being remotely controlled, although an external processor could instead be used. As such, the present invention implements what may be described as gesture syntax, rather than mere gesture semantics.
• In the event a user gesture cannot be uniquely identified by the processor unit, the system and method can display on the device choice icons showing what is believed to be the current gesture, and asking for confirmation, perhaps by the user moving a hand toward the device. The three-dimensional camera system may, but need not be, a three-dimensional TOF system. Such an imaging system readily enables hand or other object movements toward and away from the device to be reliably detected, regardless of shape or size or color or variations in ambient light, including light from the television display itself.
• Other features and advantages of the invention will appear from the following description in which the preferred embodiments have been set forth in detail, in conjunction with their accompanying drawings.
BRIEF DESCRIPTION OF THE DRAWINGS
• FIG. 1 depicts a generic prior art gesture recognition system in which device-centric coordinates are used, and hand-eye coordination and device calibration are typically required on the part of the user, according to the prior art;
• FIG. 2 depicts a user-centric gesture recognition system, according to embodiments of the present invention;
• FIG. 3A-3G depict a library of exemplary gestures, according to embodiments of the present invention;
• FIG. 4 is a flow chart depicting generic exemplary method steps implementing user-centric gesture control, according to embodiments of the present invention; and
• FIG. 5 is a flow charge depicting generic exemplary method steps implementing user-centric gesture control, including further details concerning handing off device control from one user to another user, according to embodiments of the present invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
• FIG. 2 depicts agesture recognition system100, according to embodiments of the present invention.System100 includes a preferably three-dimensional imaging system140, such as is known in the prior art, whose effective field of view encompasses a three-dimensional space in which at least one user may be present. For ease of illustration,system140 is depicted as a generic TOF system, although a non-TOF imaging system may instead be used. For example, a system employing multiple cameras might be used, as could other prior art three-dimensional imaging systems, indicated generically by140′. InFIG. 2,exemplary TOF system140 emits optical energy (shown as phantom lines emanating away from system140) towards theuser30, and detects a fraction of that optical energy reflected by the user or other object(s) within the field of view. This reflected energy is depicted as phantom lines directed back toTOF system140.Gesture recognition system100 also includes aprocessor unit150, which preferably includesmemory160, a processor ormicroprocessor170, and various input/output circuitry180. Output signals fromprocessor unit150 are coupled viapath190, to control television orother device20, responsive to user gestures that are recognized by three-dimensional imaging system100.
• As indicated by the coordinate axes, the present invention uses world coordinates that are defined relative touser30, rather than relative todevice20 orsystem100.System100 defines, relative touser30, a local environment termed herein a three-dimensional zone ofinteraction200, which zone preferably is sized to the size of the user, and will move as the user moves. User gestures, e.g., hand or arm gestures, made within the interaction zone are detected bysystem100. Preferably within zone of interaction200 a small number of three-dimensionalhot zones210,220,230 are defined, e.g., at least two and preferably three such zones. These hot zones are regions of three dimensional space defined within the larger interaction zone. In general, the user can draw shapes with one or more hands, e.g., perhaps a circle, an arc, etc., for detection within the interaction zone, However, embodiments of the present invention preferably look for occurrence of user touch gestures within the hot zones defined within the interaction zone. In such embodiments, the user can simply “touch” a hot zone, or make hand movement within a hot zone to commence or create a recognizable gesture. In practice, for a givenuser30, zone ofinteraction200 might have exemplary dimensions on the order of perhaps 3′ (1 M) in left-to-right width, 2′ (0.7 m) in height, and perhaps 2′ (0.7 m) in front-to-back depth, although other dimensions could instead be defined. Eachhot zone210,220,230 may occupy a shape, perhaps a sphere, with transverse diameter of perhaps 2′ (0.7 m) within zone ofinteraction200, although again different dimensions could be defined. However for a physicallylarger user30, all of these dimensions would automatically be scaled upward bysystem100. Thus a user with long arms would interact with a physically larger zone of interaction than a user with small arms, etc. An advantage of defining hot zones according to the present invention is to reduce occurrence of false detections, or detection of unintended gestures.
• Because the present invention uses a world coordinate system referenced touser30, it is understood that ifuser30 moves around, zone ofinteraction200 andhot zones210,220,230 can dynamically move and re-size with the current user position. Thus, if the user moves towardstelevision device20 and then makes a gesture with one or more hands, the present invention will recognize the gesture. If the user then steps backwards or sideways and makes the same gesture, the present invention will still recognize the gesture, because the gesture is made within the interaction zone, which is defined relative to the coordinate system of the user. Advantageously, a user's gestures can be recognized, whether the user is standing when making the gesture, sitting when making the gesture, as long as the user's gesture is made within the three-dimensional imaging system field of view.
• Furthermore, the velocity of the gesture may be coupled to the speed of an action on the screen ofdevice20. For instance, if the user's right hand is moved from right to left, the device can respond by shifting a sequence of images on the screen in the same direction. If the user's hand moves more rapidly, the images can shift more rapidly. If the user hand moves slowly, the images shift slowly. As noted, in the present invention velocity of the user's hand is determined in world coordinates with respect to the user. In the example of the user's hand shifting a displayed sequence of images, the coupling between the user action and device response will remain substantially constant regardless of the distance between the user and the device.
• As noted above, preferablysystem100 detects gestures that are performed inside the interaction zone, which can include regions external to the hot zones. In addition to detecting user gestures to control a device, embodiments of the present invention can also detect when a first user hands off device control to a second user.
• User30 is free to move about relative todevice20, and because (x, y, z) world coordinates are defined relative to the user, and thus zone ofinteraction200 moves with the user, to dynamically define a local environment. It is intuitive for a user to be told he or she can interact with a three-dimensional zone of interaction defined in front of him or her, centered at perhaps shoulder level. It is also intuitive for the user to be told that he or she can touch or draw shapes within three dimensional hot zones defined within the zone of interaction. As such, a user can make a gesture within the zone of interaction without having to even look atdevice20. This is easier for the user than having to concentrate ondevice20, to look perhaps for a cursor, and then to manipulate a hand to move the cursor toward a desired region on the television display. This desired elimination of hand-eye feedback further contributes to the user-friendly manner in which gestures are created, without substantial training or hand-eye coordination requirements, according to the present invention. However, as will be described, should a gesture not immediately be recognized bysystem100,system100 can temporarily display preferably threeicons240,250,260 ondevice20, which icons correspond to the preferably threehot zones210,220,230. In this embodiment, if there were N hot zones defined withininteraction zone200, then there would be a like number N of displayed icons. Of course a different number of displayed icons and a different number of hot zones could be used. InFIG. 2, one of these icons, perhapscenter icon250, can display whatsystem100 believes to be the intended gesture or perhaps a target of the gesture. The target of the gesture refers to an object on the display that is intended to respond to the gesture, perhaps an image in a sequence of images that will move in a direction and with a velocity responsive to motion of the user's hand. InFIG. 2, adjacent icons, e.g.,240,260, can give the user the option to confirm or reject the thus-interpreted gesture, e.g., by interacting with one of the preferably threehot zones210,220,230.
• It is useful at this juncture to briefly considerFIG. 3A-3G in which a library (stored or storable in memory160) of exemplary gestures is shown. InFIG. 3A, a “swiping” gesture made at head level is defined as “turn on/menu”. InFIG. 3B, a movement of both open spaced-apart hands towardsdevice20 is defined as “pause/play”. Such forward motion is readily detected byTOF camera system140. By contrast, conventional two-dimensional RGB or gray scale cameras such as used in the prior art system ofFIG. 1 have a more difficult challenge with respect to reliably detecting such movements, and with respect to not being tricked by lighting, contrast changes, background colors, etc. Gestures such as shown inFIGS. 3B and 3C would be challenging for prior art two-dimensional RGB or gray scale systems to recognize properly because of their Z-direction movement.
• InFIG. 3C, holding the open right hand briefly then moving forward can define a “select” gesture. In other embodiments, a select gesture can be defined in terms of the user's hand being present within a hot zone for a given amount of time, even without moving the hand within the zone. For example, a gesture may cause a menu to appear ontelevision device20, and the user may be given an option, perhaps an icon of a DVD if it is desired to now view a DVD movie rather than continue to watch a television show. The user can “select” to go into the DVD movie mode by selecting the DVD icon. The user could then use a gesture, perhaps as shown inFIG. 2B, to start to play the DVD movie, or if the movie is playing, to pause the movie, and so forth.
• The appearance of icons on the television device is referred to herein as a GUI event, and typically can occur to resolve potential ambiguity in a gesture. Thus, the “volume”icon250 might appear as shown inFIG. 2, to ask the user to confirm with a yes or no whether indeed the last detected gesture was intended to alter the device volume level. In such instance the user can motion left or right with a hand to select, respectively, “yes” or “no”, since the “yes” icon is to the left on the television screen, and the “no” icon is to the right. In this example, the user's will move from one hot zone to a hot zone to the left, to indicate “yes”, or will move from one hot zone to a hot zone to the right to indicate “no”. Because this gesture is responsive to an icon query on the television screen, the icon will not be interpreted as a gesture to turn on the device, or to bring up a menu, as inFIG. 3A.System100 can then confirm the user's selection, e.g., “yes”, perhaps by causing the “yes” icon to now be highlighted, or to appear to vibrate back and forth slightly. Thus, in the above embodiments, select gestures may be defined in terms of the user's hand, stationary or moving, remaining in a hot zone for a certain amount of time. It is understood that the present invention can be practiced without recourse to menus appearing on the device display, or indeed without requiring the user to view the system under control.
• InFIGS. 3D and 3E, clockwise or counter-clockwise circular motion of the open right hand commands volume change. InFIG. 3F, downward motion of both hands closely-spaced with palms downward is “mute/unmute”, while inFIG. 3G, moving both open hands toward each other defines a “turn off” gesture. Of course more or fewer gestures could be defined.
• According to embodiments of the present invention, to be unambiguously recognizable bysystem100, each gesture emphasizes syntax rather than semantics. According to the present invention, a well defined gesture has good detection and good discrimination properties. Good detection means that three-dimensionaldepth imaging system140 can readily image the gesture, and good discrimination means that the detected gesture is readily discriminated from other gestures.
• For example, inFIG. 3A, the hand swipe gesture is not readily confused with any of the other gestures shown, and thus has good discrimination properties. But if several swipes to the right meant “turn on” and several swipes to the left meant “turn off”, there would be ambiguity. These gestures would be ambiguous because after completing one swipe to the right, the hand would have to be returned from right-to-left to repeat the same gesture. But the act of returning from right-to-left might be interpreted as “turn off” rather than the return motion associated with “turn on”. This example of a poorly defined gesture also demonstrates what is meant by the preferred requirement that a good gesture be stateless or memory-less, e.g., it should not have an initial condition. Thus inFIG. 3C or3D, making continuous circular motion in a clockwise or counter-clockwise direction is stateless or memory-less because there is no requirement for a starting position. Similarly inFIG. 3F, there is no requirement that the two hands commence their downward motion from a given starting point, and thus this gesture has the desired characteristic of being stateless or memory-less. Thus, many presently preferred gesture movements might be said to be AC-coupled, unlike many DC-coupled gestures in the prior art. This use of the terminology AC-coupled or DC-coupled is analogous to an electronic signal as viewed on an oscilloscope: the DC-coupled signal will have a static DC offset, e.g., a starting point level, whereas the AC-coupled signal has no offset. However the present invention is sufficiently flexible to recognize even so-called DC-coupled gestures. Presently preferred gestures according to the present invention are not defined as sub-sets of other gestures, although they could of course be so defined. Thus if one gesture is a full circular hand rotation, perhaps to command device volume change, it could be potentially confusing to also have a different gesture that is a half-circle, e.g., a sub-set of the full circle gesture.
• According to the present invention, gestures are disambiguated by assigning what might potentially be an ambiguous gesture to a single function, e.g., the swipe gesture ofFIG. 3A could be ambiguous if a swipe to the right defined one gesture, and a swipe to the left defined another gesture. However the back-and-forth swipe gesture ofFIG. 3A defines a single gesture, here turn-on/menu. Note too that gestures should not include motions that a user might normally make, for example, applauding, scratching one's nose or head. These movements can occur almost subconsciously and should not be used to define gestures, according to the present invention.
• In many instances the user may be seated on a chair or couch, with another person or persons sitting close by. Understandablysystem100 will define a single zone of interaction adjacent the user, and will ignore motions, including gesture-like motions, made by other persons. However if the user wishes to transfer remote control overdevice20 to another person, a gesture may be defined to alertsystem100 that a user in a different location, perhaps seated to the right of the original user, is about to take control.System100 would then redefine the zone of interaction and the hot spots within, relative to the new user. At this juncture, motions including gesture type motions made by the original user will not be responded to, because they no longer occur within the relevant zone of interaction.
• FIG. 4 depicts an exemplary algorithm such as can be stored inmemory160 in system100 (seeFIG. 2) for execution by processor170 (or other processor) to implement gesture recognition, according to embodiments of the present invention. Initiallysystem100 will acquire several frames of data to imageuser30, and to then define aninteraction zone200, and hot spots orhot zones210,220, etc. therein for the imaged user. Atmethod step300, the user makes a “start” or “turn-on” gesture, for example as shown inFIG. 3A, and atstep310,system100 acquiresuser20. Atstep320,system100 tracks the imaged user, which means that if the user moves about relative to the position of the controlleddevice20,system100 can successfully image the user. Gesture recognition according to the present invention is user-centric, not device-centric, and the user is free to move about relative to the position ofdevice20.
• Atstep340, the system confirms that the imaged user is indeed being tracked. If for some reason tracking is not occurring, an “ACQUIRE” command is used atmodule330 and step310 is repeated until the user is properly acquired and tracked.
• Once user tracking is confirmed,method step340 passes off to step350, and appropriately sized user-centric zones including hot zones are defined for this user. As noted, a physically large user will have larger volume zones defined than would be the case for a physically small user. Acquired images ofuser20 enablesystem100 to approximate the user's size and to causestep350 to generate appropriately sized three-dimensional zone spaces in which gesture detection preferably will occur.
• Atmethod step360, acquired images of the user and user hands within the detection zones and hot zones are examined to detect user made gestures. As noted earlier, the preferred use of hot zones in which touch type gestures should be made tends to reduce false detection of gestures, as well as ambiguous identification of detected gestures.
• Atmethod step370,system100 compares the detected gesture with contents of a user gestures that have been previously stored, e.g., inmemory160 insystem100. A best determination of the user gesture is made, and this gesture is then mapped to appropriate command(s) for controlleddevice20. For example, if the user gesture is determined to mean “increase volume level” fordevice20, amethod step370 the appropriate electronic command signals to cause an increase indevice20 audio volume level will be generated. These signals can be coupled todevice20 via cable(s), or wirelessly, e.g., IR, Bluetooth, etc.
• Having made the appropriate parameter change to operation ofdevice20, atmethod380,system100 will continue to image the user. If a predetermined amount of time lapses without detection of further gestures,system100 will look to see whether there is a new, substitute, user, perhaps positioned other than precisely where the present-former user was last positioned. If a new user is found, and it is determined that the new user is performing an “acquire me” gesture, the software inmemory160 being executed, e.g., by processor170 (or other processor) will substitute the new user for the old user. But if no new user is found, the old user will remain as the current user, whose continuing user gestures will be imaged and processed to controldevice20.
• FIG. 5 depicts a further exemplary algorithm such as can be stored inmemory160 in system100 (seeFIG. 2) for execution by processor170 (or other processor) to implement gesture recognition, according to embodiments of the present invention, and more particularly to implement a change in user control ofdevice20. As was the case withFIG. 4,system100 initially acquires several frames of data to imageuser30, and to then define aninteraction zone200, andhot spots210,210, etc. therein foruser20.
• Similarly to what was described with respect toFIG. 4, atmethod step400, the user makes a “start” or “turn-on” gesture, and atmethod step410,camera system140 acquires a frame of three-dimensional data that includes any user interaction withininteraction zone200. Atstep420system100 searches for an acquisition gesture. As was described with respect toFIG. 4, the gesture preferably is identified by comparing gestures stored inlibrary160 with a just-acquired gesture, discerned for example byprocessor170 from data output fromcamera system140. If an acquisition gesture is not acquired, the routine returns to step410, to examine a next acquired frame of camera system data. Eventually a acquisition gesture will be acquired atstep420.
• At thisjuncture step430 is bypassed and atstep440,system100 examines the hot spots orhot zones210,220,230, which act as segments of theinteraction zone200, to determine whether a touch gesture has been detected. If detected, the user interface (UI) signals a “touch” or “hold” event, and the method branches to step460 to test whether tracking of the user has been lost. Input to step460 also includes the absence of touch detection atstep440. If tracking has not been lost, step460 branches todetection step510. If a gesture is detected, the method branches to step520 and a user interface “command name” signal is issued. Step530 tests to determine whether a stop command has been detected, e.g., a stop gesture. If a stop command has been detected, then atstep540system100 will turn-off the application, in this example,television20. But iftest530 determines that the gesture is not a stop command, then the routine branches back to step530, to examine the next acquired frame of data fromcamera system140.
• Detection step510 also outputs to step550, which tests for the presence of another user. Perhaps the first user has an errand to run and will pass control ofdevice20 to a second user. As noted, one can define a gesture intended to alertsystem100 to the transfer of remote control command from one user to another user, perhaps a second person sitting next to the initial user. If it is not apparently time to search for another user, then step510 branches back to step530, and the next frame of acquired data is examined. However ifstep510 determines that no gesture is detected and it is time to look for a new user, the routine branches to step480. If a new user cannot be determined, the routine branches back to step430, and data from the next acquired frame is analyzed. However ifstep480 determines a new user is present, then the routine branches to step490 and a signal UI acquisition gesture is tested. Step500 then retargets tracking.
• In summary, it will be appreciated that user-centric gesture recognition provides many advantages over prior art techniques. Advantageously, embodiments using hot zones can reduce incidents of false gesture identification and recognition. The present invention may be practiced with a variety of prior art three-dimensional imaging systems140, including without limitation Canesta, Inc. time-of-flight (TOF) systems. An advantage of such TOF systems is that overall system and method reliability can be enhanced, as gesture recognition is not likely to be tricked by user clothing colors, reflections, by ambient light, and the like.
• Modifications and variations may be made to the disclosed embodiments without departing from the subject and spirit of the invention as defined by the following claims.

Claims (18)

1. A method implementing user-centric gesture recognition enabling a user to remotely gesture-control a device, the method comprising:

disposing a three-dimensional imaging system whose field of view encompasses a three-dimensional space in which at least one user may be present, and recognizing a user when said user is within said field of view;

providing a gesture recognition system that defines a world coordinate system within said three-dimensional space based upon location of said user as determined by said imaging system, said world coordinate system moving with movement of said user;

said gesture recognition system defining relative to a current position of said user as determined by said imaging system a dynamically-sized and positioned three-dimensional zone of interaction within which zone said user may make at least one gesture;

said gesture recognition system storing a gesture library of at least one pre-defined gesture useable to control said device;

said gesture recognition system recognizing a gesture made by said user within said three-dimensional space and comparing said gesture to contents of said gesture library;

upon finding a match between a recognized said user-gesture and contents of said gesture library, outputting a signal useable to control said device.

2. The method ofclaim 1, further including defining at least one sub-region within said three-dimensional zone of interaction, wherein presence of a portion of said user can be detected.

3. The method ofclaim 1, wherein said gesture library stores at least one gesture selected from a group consisting of (i) a gesture signifying desire to acquire control of said device, (ii) a gesture signifying handing off remote control of said device from said user to a successor user, (iii) a gesture made with one hand, (iv) a gesture made with two hands, and (v) a gesture causing a menu to be displayed on said device.

4. The method ofclaim 1, further including causing said device to alert a user that a user-gesture may not have been correctly recognized by said gesture control system, and inviting said user to confirm improper recognition of a gesture.

5. The method ofclaim 4, wherein a user can confirm improper recognition of a gesture by making a user movement within a sub-region defined within said three-dimensional zone of interaction.

6. The method ofclaim 5, further including causing said device to signal said user that improper recognition of a gesture may have occurred, and causing said device to provide an option for said user to at least confirm said improper gesture recognition.

7. The method ofclaim 1, wherein said gesture library stores gestures that are stateless.

8. The method ofclaim 1, wherein said three-dimensional imaging system is a system selected from a group consisting of a (i) parallex-method imaging system, (ii) a time-of-flight imaging system, (iii) a phase-based time-of-flight imaging system, (iv) a shutter-based time-of-flight imaging system, and (v) a system using at least two spaced-apart cameras.

9. The method ofclaim 1, wherein said device includes at least one of (i) a display, (ii) a computer screen, (iii) a control panel, and (iv) a television receiver.

10. A system enabling user to remotely gesture-control a device using user-concentric gesture recognition, the system comprising:

a three-dimensional imaging system whose field of view encompasses a three-dimensional space in which at least one user may be present, and recognizing a user when said user is within said field of view;

a gesture recognition system, coupled to said three-dimensional imaging system, that defines a world coordinate system within said three-dimensional space based upon location of said user as determined by said imaging system, said world coordinate system moving with movement of said user;

said gesture recognition system including means for defining, relative to a current position of said user as determined by said imaging system, a dynamically-sized and positioned three-dimensional zone of interaction within which zone said user may make at least one gesture; and

a gesture library of at least one pre-defined gesture useable to control said device;

said gesture recognition system recognizing a gesture made by said user within said three-dimensional space and comparing said gesture to contents of said gesture library;

upon finding a match between a recognized said user-gesture and contents of said gesture library, said gesture recognition system outputting a signal useable to control said device.

11. The system ofclaim 10, wherein said gesture recognition system defines at least one sub-region within said three-dimensional zone of interaction, wherein presence of a portion of said user can be detected.

12. The system ofclaim 10, wherein said gesture library stores at least one gesture selected from a group consisting of ((i) a gesture signifying desire to acquire control of said device, (ii) a gesture signifying handing off remote control of said device from said user to a successor user, (iii) a gesture made with one hand, (iv) a gesture made with two hands, and (v) a gesture causing a menu to be displayed on said device.

13. The system ofclaim 10, further including means for causing said device to alert a user that a user-gesture may not have been correctly recognized by said gesture control system, and inviting said user to confirm improper recognition of a gesture.

14. The system ofclaim 13, wherein a user can confirm improper recognition of a gesture by making a user movement within a sub-region defined within said three-dimensional zone of interaction.

15. The system ofclaim 14, further including means for causing said device to signal said user that improper recognition of a gesture may have occurred, and causing said device to provide an option for said user to at least confirm said improper gesture recognition.

16. The system ofclaim 10, wherein said gesture library stores gestures that are stateless.

17. The system ofclaim 10, wherein said three-dimensional imaging system is a system selected from a group consisting of a (i) parallex-method imaging system, (ii) a time-of-flight imaging system, (iii) a phase-based time-of-flight imaging system, (iv) a shutter-based time-of-flight imaging system, and (v) a system using at least two spaced-apart cameras.

18. The system ofclaim 10, wherein said device includes at least one of (i) a display, (ii) a computer screen, (iii) a control panel, and (iv) a television receiver.

Images